Applied Intelligence: The New Breed of Business Application
The large volumes of business data now generated by electronic transactions can be converted, by professionals using business intelligence (BI) tools, into valuable commercial insights. The Market research organisation, International Data Corporation estimates that the market for business intelligence tools alone will reach $5.2 billion by 2003. Certain financial services firms may process 80 Million transactions per year and store dozens of terabytes, growing by as much as 100 percent a year.
Benefits of BI
Business intelligence can be used to answer fundamental questions within organizations such as:
- What are my most popular products?
- How are promotions affecting our sales?
- Which are our best suppliers?
- Why have our Northern customers stopped visiting the website?
- Where are my most loyal customers?
Business intelligence applications look at the information needed to answer these questions and then search the data needed to generate that information. The resulting answers can be optimally targeted throughout an organization via an interface allowing users to explore the data from different perspectives.
Business intelligence products are often used to retrieve and analyze the contents of data warehouses. Used correctly, BI can also help spot patterns in data that normally would not be apparent. The insights provided by BI make certain forms of marketing possible e.g. defining personalization rules and measuring their effectiveness. Other enabled techniques include "events-based" operation-a company, for example, can stop running an e-mail campaign the moment stocks start to run out.
History of BI
BI tools first existed a few years ago in two basic forms: Executive Information Systems (EIS) and Decision Support Systems (DSS).
EIS tools created applications that could be used by business managers to analyze data to support decision making. EIS systems typically ran on mainframes, cost e.g. over $100,000 and were used by only top-level managers. DSS tools, which were available to more technical knowledge workers, supported advanced analytical techniques and were comparatively cheap.
The rise of client/server and Internet computing and the move towards autonomous business units, made these systems inappropriate. BI grew up in the 1990s as companies began wondering what to do with the huge quantities of data they were accumulating from their ERP systems, call centers and the Internet. Business intelligence solutions are now being employed as hosted solutions -often with a portal interface, and are used increasingly for analysis of data collected from e-commerce and packaged application sources.
From a developer's point of view, some very effective BI technologies are not necessarily leading-edge but to get the most benefit from them, companies have to adopt them effectively. There are several component technologies constituting a business intelligence solution within a firm:
- Datamarts, data warehouses
- ETL (Extraction, Transformation, and Loading) helps integrate heterogeneous enterprise systems and applications, so that data may be efficiently moved across the enterprise and loaded into corporate databases.
- Reporting, query, on-line analysis, visualisation
- Decision modelling
- Data mining (spotting connections between bits of apparently unrelated data, using e.g. expert systems, artificial intelligence and fuzzy logic )
- Decision Support and Executive Information Systems (EIS e.g. portals, 'dashboards' and 'scorecards')
- Multi-Dimensional Analysis and OLAP (online analytical processing)
- Statistics and Technical Data Analysis (inc trend analysis, forecasting)
- Competitive Analysis
The purpose of business intelligence tools is to allow users to access a graphical/ numerical summary of the state of their business. Unfortunately, in relational database management systems, data is structured in rows and columns that run across multiple tables. Also, the data are highly normalized to prevent redundancy, ensure integrity, and allow the data to be accessed flexibly. Data in this form are useless to business users.
There are two different approaches to presenting users with information in a useful format. The first involves multidimensional storage of data in 'hypercubes'. The second method requires normalized data structures to be 'reconstructed' into a "view" that maps directly back to their original business representation.
Multidimensional databases (MDBMS) store data in matrices along related attributes, or dimensions, such as product, cost, sales, etc. Most systems use complicated indexing and compression algorithms to reduce the size and optimize access to the data. Other systems write the data directly to disk as multidimensional arrays.
An alternative approach is to provide users with multidimensional 'logical views' of relational data. This is accomplished by combining tables that are commonly linked into a single logical table, and then mapping table and field names to more user-friendly equivalents.
The most common type of BI navigational functionality, 'slice-and-dice', allows users to view data across any dimension or combination of dimensions within a hypercube. All BI tools also, of course, allow users to "drill down" through layers of data to more detailed information.
For the first time, these tools together enable closed-loop decision making processes that attempt to automate much of the understand-act-evaluate cycle. Advanced analytics can enable rapid refinement of business processes and even allow qualitatively different management practises (e.g. metrics-driven management). These and other advances in BI will form the subject of my next article.
Patrick Andrews is managing director of break-step productions, a consultancy firm specializing in designing digital businesses. His areas of interest include interactive marketing, machine intelligence and software design. Contact him at firstname.lastname@example.org.